Risks of AI in Event Management: Why AI Needs Support

Long before delegates arrive, an event can become difficult to control in ways that are easy to miss.

A speaker has not completed registration. An abstract still needs the right reviewer. A payment does not match its invoice, and moving one session creates three new clashes. None of these jobs is unusual, but together they can consume a team’s attention.

The risks of AI in event management become clear when a tool is asked to take responsibility for all these connected jobs. A single incorrect assumption can move from an abstract or registration record into a payment decision, reviewer assignment or published schedule.

It is easy to understand why artificial intelligence looks attractive at this point. Give it the data, explain the event and let it take care of the repetitive work. In theory, an AI assistant could read submissions, recommend reviewers, answer registration questions, flag payment issues and build a schedule in minutes.

AI can certainly help with parts of that work. It can turn notes into a first draft, summarise common questions, suggest email wording and help a team find patterns in feedback. However, helping with a task is very different from taking responsibility for the whole event.

An event is not a single conversation with a chatbot. It is a chain of connected commitments involving authors, reviewers, speakers, attendees, sponsors, venues, payment providers and organising committees. A change in one place often creates consequences somewhere else. If an accepted speaker withdraws, the team may need to update the session, inform the chair, revise the website, contact a reserve presenter and check whether a refund is due. A general AI tool may produce a confident answer, yet still miss one of those links.

That is why AI should not become the operating centre of an event. It works better as an assistant sitting beside a structured event management platform, with people remaining responsible for important decisions.

The risks of AI in event management start with false confidence

Some errors announce themselves immediately, while others arrive in language that feels polished and dependable. The more fluent the answer sounds, the easier it becomes to overlook the need for verification.

Confidence is not the same as accuracy

The most obvious risk is also the easiest to underestimate. Generative AI is designed to produce a useful-looking response. That response can be clear, polished and completely believable, even when a detail is incorrect.

The US National Institute of Standards and Technology calls this problem “confabulation”. Its guidance describes it as confidently presented false or incorrect content. NIST also warns that these systems can produce invented logic or citations that make a wrong answer appear more trustworthy. This becomes especially risky when AI is used in decisions that have real consequences. (NIST Generative AI Profile)

The organiser remains responsible for the answer

This is not merely a theoretical concern. In Moffatt v. Air Canada, a customer relied on incorrect bereavement-fare information provided by the airline’s chatbot. The British Columbia Civil Resolution Tribunal held the airline responsible for the misinformation. The lesson for event organisers is straightforward: if an automated assistant gives an attendee the wrong refund, ticket or cancellation information, the organisation cannot simply blame the tool. (American Bar Association summary of the decision)

Imagine an attendee asking, “Can I cancel after the early-bird deadline and still receive a full refund?” The AI may find an old policy, confuse one ticket type with another or overlook an exception agreed with a sponsor. Its answer may still sound perfectly reasonable. By the time a staff member discovers the mistake, the attendee has acted on it and expects the organisation to honour it.

A reliable event platform approaches this differently. The refund policy is attached to the relevant ticket, order and payment record. Staff can see what was purchased, when it was purchased and which rule applies. If an exception is made, it can be recorded against the attendee’s account. AI may help explain that policy in plain language, but it should not invent or choose the policy.

Abstract management involves more than reading text

An abstract moves through several hands before it becomes part of the final programme. Reading its content is only one step in a process filled with deadlines, revisions, permissions and decisions.

Every submission follows a process

Abstract management looks like an ideal job for AI because so much of the material is written. An AI tool can summarise a proposal, extract keywords and suggest a topic. Those are genuinely useful abilities. The trouble begins when the tool is expected to manage the entire submission process.

Every conference has its own rules. One event accepts individual papers and panels. Another accepts posters, workshops and roundtables. Word limits may vary by submission type. Some fields are visible to reviewers, while author names and affiliations must remain hidden. A presenter may appear on several submissions but can only speak twice. A revised abstract may replace an earlier version after the deadline with the chair’s approval.

A conversation is not an official record

A conversation-based tool can interpret each request as it arrives, but it does not automatically provide a dependable record of which version is final, who approved an exception or whether every required field was completed before the deadline. If the organiser has to check a spreadsheet after every AI-assisted action, very little time has been saved.

A purpose-built abstract system provides a fixed path. Authors submit through the correct form. Required information is collected consistently. Deadlines and submission statuses are visible. Changes are linked to the right record. Organisers can then use AI for supportive work, such as suggesting keywords or creating a short internal summary, without asking it to become the official home of the submission.

Use case: 2,000 submissions arrive in different formats

Consider an international academic conference receiving 2,000 abstracts. Some authors submit individual papers, others submit panels with several speakers, and a few request confidential corrections after the deadline.

If the team relies only on AI and shared files, it must trust the system to identify every author, preserve every revision, apply different rules to different formats and maintain the correct status for each submission. One mistaken identity or overwritten version can affect acceptance letters, registration checks and the final programme.

With event management software, each submission has its own record, history, authors, files and status. AI can assist with tagging or summaries, while the platform preserves the process. That division of labour is both faster and easier to defend when someone questions a decision.

Peer review requires confidentiality, context and human judgement

Reviewers do more than judge the quality of a proposal. Their decisions influence whose work is heard, how the programme develops and whether authors trust the conference process.

Good reviewing goes beyond scoring the writing

Reviewing abstracts is not simply a matter of ranking good writing. Reviewers consider relevance, originality, evidence, methods, fit with the conference theme and the value a proposal may bring to the audience. They also notice conflicts of interest and disciplinary differences that are difficult to capture in a general prompt.

Confidential work should remain confidential

Confidentiality is another concern. Abstracts can include unpublished findings, original ideas, personal information or commercially sensitive work. Uploading that material into an external AI tool without checking how it is handled may breach conference rules or the author’s reasonable expectations.

The National Institutes of Health prohibits scientific peer reviewers from using generative AI to analyse or write critiques of grant applications and proposals. Its notice explains that uploading application content or original concepts to online AI tools violates its confidentiality and integrity requirements. A conference is not governed by NIH policy unless that policy directly applies, but the concern is highly relevant to academic event review. (NIH notice on generative AI in peer review)

Final decisions need context and accountability

There is also the question of fairness. An AI system may favour familiar language, conventional structures or topics that resemble material it has seen before. It may undervalue work from early-career researchers, less represented regions or disciplines with different writing traditions.

Even if the recommendation is only a score, people can begin to accept it without asking how it was reached. The safer approach is to let the review committee define the criteria and let qualified reviewers make the assessment. Software can distribute submissions, protect blinded information, collect scores, send reminders and highlight missing reviews. AI may help an organiser identify possible subject matches, but a human should confirm assignments, conflicts and final decisions.

Use case: a promising abstract receives an unexpectedly low score

Suppose an AI system gives a low rating to a proposal written in imperfect English. The research is strong, but its value is not expressed in the polished style the system tends to reward. If the event automatically rejects the lowest-scoring submissions, the work may disappear without a subject expert ever reading it.

In a structured review process, several reviewers can assess the proposal against published criteria. A track chair can investigate a large difference between scores and record the final decision. That process does not eliminate human bias, but it makes responsibility visible and gives organisers a way to review unusual outcomes.

Registration is full of exceptions that matter

The registration desk begins working long before attendees arrive at the venue. Every ticket choice, discount and special request adds another detail that must remain accurate throughout the event.

A registration form carries many connected rules

Registration appears simple until real attendees begin using it. There are early-bird rates, member prices, student tickets, invitation codes, group bookings, workshops, meal choices, accessibility requests, visa-letter requirements and tax details. One person may be an author, reviewer, speaker and paying attendee at the same time.

An AI assistant can answer common questions, but it should not be expected to create or maintain the official registration record. If it misunderstands an attendee’s role, applies the wrong discount or confirms a sold-out workshop, the problem spreads to capacity planning, revenue reports, badges and the schedule.

One attendee needs one dependable record

A structured registration system checks the choices available for a particular ticket and stores the answers in the attendee’s record. It can stop sales when capacity is reached, apply an approved code and show the organiser who has paid. Most importantly, the same information remains available to the finance, programme and check-in teams.

Use case: an accepted speaker has not completed registration

An AI tool may read the acceptance list and registration export, then conclude that “Dr A. Sharma” and “Anita Sharma” are different people. Alternatively, it may merge two attendees who happen to share a name. Either error can lead to an unnecessary reminder or, worse, an accepted talk remaining in the programme even though the presenter never registered.

An event platform links the person, submission, acceptance status and registration activity within one controlled record. Staff can review uncertain matches instead of letting a language model guess. AI can draft the reminder, while the platform determines who should receive it.

Payments cannot depend on a plausible answer

A payment confirmation changes what an attendee and the organising team expect from each other. Once money moves, estimates and assumptions must give way to exact, traceable records.

Financial records must match what actually happened

Money leaves little room for improvisation. Attendees expect the amount charged, invoice issued, tax recorded and refund processed to match. Finance teams also need a clear history of discounts, partial payments, cancellations, failed transactions and manual adjustments.

AI can help explain a charge or draft a payment reminder. It should not independently decide whether a payment succeeded, whether an attendee owes tax or whether a refund should be issued. Those facts must come from the order, invoice and payment provider.

Outsourcing does not remove responsibility

Payment handling also carries formal responsibilities. The Payment Card Industry Security Standards Council states that even merchants outsourcing all payment processing remain responsible for ensuring that providers protect account data and meet the relevant requirements.

A general AI assistant is not a substitute for a properly managed payment process. (PCI Security Standards Council guidance)

Use case: a group booking includes a partial refund

A university buys ten registrations, adds two paid workshops and later replaces one attendee. One workshop is cancelled, so part of the order must be refunded. The revised invoice must still show what was purchased, what changed and what remains payable.

An AI tool may calculate the amount correctly in one conversation, yet fail to update every related record. It may also lack the authority to confirm the actual payment status. Event software keeps the order, invoice, payment and refund connected. A staff member can approve the adjustment, and the record remains available if the university’s accounts team asks about it months later.

Scheduling is where small mistakes become public problems

Attendees experience the programme as a clean timetable of rooms, speakers and sessions. Organisers know that this simplicity is created by resolving countless dependencies behind the scenes.

A timetable is a web of promises

A conference schedule is a web of promises. A speaker cannot be in two rooms at once. A session chair may have another commitment. A workshop needs the largest room. A sponsor may have a contracted speaking slot. A virtual presenter needs a suitable time zone. Two sessions aimed at the same audience should not compete if that can be avoided.

AI can produce an impressive-looking timetable, especially when given a clean list of sessions. However, real event data is rarely clean or final. Speakers withdraw, rooms change, sessions run longer than expected and organisers make exceptions that were agreed in emails or meetings. Unless every condition is recorded and checked, a polished schedule can hide serious clashes.

Approved changes must reach the whole event

This is where event scheduling software earns its place. It creates sessions from accepted submissions, attaches the correct speakers and rooms, checks known conflicts and publishes approved changes to the attendee-facing programme. The organiser can still make judgement calls, but the system keeps the connected records together.

Use case: one withdrawal affects five parts of the event

A panellist withdraws two days before the conference. Replacing that person may change the session description, moderator notes, speaker page, room setup and attendee notification. If those items live in separate AI conversations or documents, one will probably be missed.

In an event platform, the team updates the underlying speaker or session record and can see where that information appears. AI can help rewrite the session description and notification. However, the platform controls where the approved information is published.

Personal data needs clear boundaries

People provide personal details because they trust the event team to use them carefully. That trust can be weakened quickly when information is copied into tools without a clear reason or adequate safeguards.

Events collect more information than we realise

Events collect more personal information than organisers sometimes realise. Registration forms may include contact details, employer information, dietary choices, accessibility needs, travel plans and payment-related records. Abstract systems may contain affiliations, biographies and unpublished research. Networking tools add messages, meeting requests and personal schedules.

Putting all of this into an AI tool because it is convenient creates a new question: where is the information going, who can access it and how long will it remain there?

Use only the data a task genuinely needs

The UK Information Commissioner’s Office advises organisations using AI with personal data to assess risks, act lawfully and transparently, limit the data used and maintain accountability. It also notes that AI can increase existing risks or make them harder to manage. (ICO guidance on AI and data protection)

Event organisers should therefore avoid copying entire attendee lists, confidential abstracts or payment records into public AI tools. Before using any AI service, the team should understand its data terms, permissions and retention settings. It should also decide which information the tool genuinely needs. Often, a limited or anonymous extract is enough for the task.

Purpose-built software gives the event a dependable memory

When hundreds of actions happen across several months, memory cannot depend on inboxes, chat histories or individual team members. A dedicated platform preserves the relationships between people, decisions and transactions as the event changes.

Structure matters more than a clever response

The strongest argument for event management software is not that it is more fashionable or more intelligent. It is that it gives the event a dependable memory.

A platform such as Dryfta connects submissions, reviews, contacts, registrations, orders and schedules through defined records and permissions. According to Dryfta’s product information, organisers can use automated or manual reviewer assignments, configurable review processes, registration forms, online and offline payment records, invoices, purchase orders and conflict-aware scheduling. (Dryfta event management platform)

Comparable event platforms follow the same broad principle, although their features and strengths differ. They are designed around event workflows. They know the difference between an abstract and a review, an order and a payment, a session and a room. More importantly, they allow organisers to decide who can view, change and approve each item.

Connected systems become more valuable as events grow

Dryfta’s published case study for the 2023 International Maternal and Newborn Health Conference offers a useful example of scale. The event reportedly managed 2,722 abstracts, more than 4,000 reviews, over 2,700 attendees and more than $500,000 in ticket sales. It also recorded 24,990 additions to personal schedules. These are vendor-reported figures, so they should be read as a customer case study rather than independent research. Even so, they illustrate why one connected system matters when submissions, reviews, registration, payments and schedules all affect one another. (Dryfta IMNHC 2023 case study)

The sensible model is AI plus event software plus people

AI and event software solve different kinds of problems, so asking either one to do everything creates unnecessary risk. Human judgement connects them by deciding when speed is useful and when care must take priority.

Give each part of the team the right job

Understanding the limitations of AI in event management does not mean rejecting AI. Used carefully, it can remove a surprising amount of repetitive work.

An event team might use AI to:

• Turn an organiser’s notes into a first draft of an email;

• Summarise recurring themes in post-event feedback;

• Suggest keywords for abstracts, subject to human checking;

• Produce a plain-language version of an approved policy;

• Draft session descriptions from confirmed information;

• Suggest answers for a help centre that staff review before publication; and

• Prepare reminder wording for lists created by the event platform.

At the same time, the event platform should remain the official source for:

• Submission versions and acceptance statuses;

• Reviewer assignments, conflicts and completed reviews;

• Registration choices and attendee roles;

• Orders, invoices, payments and refunds;

• Rooms, speakers, sessions and published schedules; and

• Permissions, approvals and change history.

People should remain responsible for acceptance decisions, unusual refunds, policy exceptions, sensitive communications and final programme choices. NIST’s broader AI Risk Management Framework emphasises reliability, safety, accountability, transparency, privacy and human-centred risk management. Those principles fit event operations remarkably well. (NIST AI Risk Management Framework)

A practical checklist before using AI for an event task

Good safeguards begin before information is uploaded or an automated answer is accepted. A few clear checks can reveal whether a task is suitable for AI and where human approval is still needed.

Six questions to ask before you proceed

Before handing a task to AI, ask six simple questions.

1. Does this task involve confidential or personal information? If it does, check whether the information can be removed, reduced or kept inside an approved system.
2. Would a wrong answer affect someone’s money, eligibility or reputation? If yes, require a human decision and verify the underlying record.
3. Is the AI reading the latest approved information? Do not assume it knows about a policy change, withdrawal or private exception.
4. Where will the final decision be recorded? Important outcomes belong in the event platform, not only in a chat history.
5. Can a person explain and correct the result? There should always be a clear route for review.
6. What happens if the AI is unavailable? The team should still be able to find registrations, payments, reviews and schedules in the main event system.

These questions do not slow a team down. They prevent small conveniences from becoming large problems later.

Final thought

Every shortcut reveals what an organisation values when time is limited and pressure is high. Strong event operations protect accuracy and trust even while teams look for faster ways to work.

Let AI assist without giving away control

Event management is built on trust. Authors trust organisers with unpublished work. Reviewers trust them to protect confidentiality. Attendees trust registration details and payment records.

Speakers trust the published programme. Sponsors trust agreed commitments will be honoured. AI can support that work, but it cannot carry the responsibility on its own. It may draft quickly, find patterns and reduce routine writing. It can also misunderstand context, expose sensitive information or produce an answer that sounds far more certain than it should.

Managing the risks of AI in event management therefore requires clear boundaries. AI can assist with suitable tasks, but verified event records and accountable people must remain at the centre of every important decision.

A good event management platform such as Dryfta, or another purpose-built alternative suited to the event, provides the structure that AI lacks. It keeps records connected, applies established rules and gives organisers control over approvals and changes. Human judgement then handles the moments that require fairness, context and care.

The best future for event management is not a choice between people and AI. It is a well-run partnership in which AI assists, event software organises and people remain accountable.

Frequently asked questions

Event teams often ask where AI genuinely saves time and where it introduces new risks. The answers below address the practical concerns that tend to arise when organisers consider using it in daily operations.

What are the main risks of AI in event management?

The main risks include inaccurate answers, mishandled confidential information, incorrect attendee records, unfair abstract assessments, payment errors and scheduling conflicts. These problems become more serious when AI output is accepted without checking the event’s approved policies and official records.

Can AI manage an event on its own?

AI can complete or support individual tasks, but it should not manage an entire event without structured software and human supervision. Abstracts, reviews, registrations, payments and schedules are closely connected. A mistake in one area can quickly affect several others.

Important decisions and official records should therefore remain inside an event management platform.

Where can AI safely help event organisers?

AI is useful for drafting emails, summarising feedback, suggesting abstract keywords, preparing session descriptions and turning approved policies into clearer language. Staff should still check the output before it is sent or published, particularly when it concerns money, eligibility, confidential information or programme changes.

Why use event management software if AI can automate tasks?

AI generates responses, while event management software maintains the records and rules that keep an event organised. A purpose-built platform connects people, submissions, reviews, tickets, payments and sessions. It also gives team members appropriate access and provides a consistent place to record approvals and changes.

Can AI review conference abstracts?

AI can help identify themes or suggest possible subject categories. However, it should not make acceptance decisions on its own. Abstract review requires subject knowledge, confidentiality, conflict checks, published criteria and human judgement. Qualified reviewers and programme chairs should remain responsible for scores and final decisions.

Is it safe to upload attendee or abstract data to an AI tool?

Not automatically. Organisers should first check what data the AI provider collects, where it is stored, who can access it and whether it may be used for other purposes. Confidential abstracts, payment details, accessibility information and complete attendee lists should not be uploaded to an unapproved AI service. When AI is genuinely useful, remove names and unnecessary personal details wherever possible.

How does Dryfta help organisers manage connected event tasks?

Dryfta brings abstract submissions, reviewer assignments, registrations, payments, invoices, contacts and programme scheduling into one platform. This gives organisers a dependable record of what has happened and what still needs attention. Teams can then use approved AI tools for supportive tasks without making a chatbot responsible for the event’s official data or decisions.

What should organisers look for in event management software?

Look for software that fits the event’s actual workflow. Useful capabilities may include configurable submission and review forms, manual and automated reviewer assignment, registration rules, payment and invoice records, conflict checking, schedule publishing, staff permissions, reporting and responsive support. The right choice depends on the type, size and complexity of the event.

Give AI a supporting role while Dryfta keeps the event organised

Reducing AI risk does not require giving up automation or returning to manual spreadsheets. It starts with a reliable event system that gives every helpful tool accurate information and clear boundaries.

Bring the entire event workflow into one connected platform

AI is most useful when your event information is already accurate, organised and easy to verify.

Dryfta provides that foundation by connecting abstract submissions, reviews, registration, payments, attendee records and programme scheduling in one place.

Instead of asking an AI assistant to remember every exception or piece together information from several files, your team can work from clear records and established processes. AI can still help with writing, summaries and routine communication, while Dryfta keeps the event data connected and your organisers remain in control.

If your team is spending too much time reconciling spreadsheets, checking disconnected systems or correcting preventable mistakes, it may be time to see how a purpose-built platform can help.

Schedule a personalised Dryfta demo and explore a more dependable way to manage your next event.